Power distribution terminal equipment based on edge calculation and multi-mode perception, power distribution network inspection method, equipment and medium

Abstract

The invention discloses power distribution terminal equipment based on edge calculation and multi-mode perception, a power distribution network inspection method, equipment and a medium, and relates to the technical field of automatic power distribution. Therefore, active early warning can be carried out through the edge computing capability at the initial stage of equipment abnormity, the real-time performance and reliability are good, and the state monitoring of the power equipment is ensured to be supported. The power distribution terminal equipment comprises a regulation and control module and an Internet of Things module; the regulation and control module acquires telemetry data of the power distribution network regularly and transmits the telemetry data to the Internet of Things module through a preset private protocol between the regulation and control module and the Internet of Things module; and the Internet of Things module receives and stores the telemetry data transmitted by the regulation and control assembly, receives a real-time image of the power distribution network acquired and uploaded by an externally accessed camera assembly in real time, carries out research and judgment processing on the real-time image, generates a patrol inspection result of the power distribution network in combination with the telemetry data, and transmits the patrol inspection result to an information management area for processing.

Description

Technical Field

[0001] This application relates to the field of automated power distribution technology, and in particular to a power distribution terminal device, power distribution network inspection method, equipment and medium based on edge computing and multimodal sensing. Background Technology

[0002] Distribution automation, as a core component of building a robust smart grid, is a key technology for ensuring the safe and stable operation of the power grid by enabling real-time monitoring and remote control of distribution network equipment through distribution terminals. Traditional distribution terminals (such as DTUs and FTUs) are based on "three-remote" (telemetry, telesignaling, and telecontrol) technology, which can collect electrical parameters such as current, voltage, and switch status and achieve remote control, laying the foundation for distribution network automation. However, with the increasing demands for power grid intelligence, monitoring methods relying solely on electrical quantity data are no longer sufficient to meet the requirements for comprehensive perception of equipment status and risk prediction.

[0003] In related technologies, to expand monitoring dimensions, a common approach is to deploy independent video surveillance systems and power distribution automation systems in parallel. Specifically, ordinary security cameras are installed in the power distribution room, transmitting video streams to a video surveillance platform via a separate network for security purposes (such as theft and vandalism prevention). Simultaneously, DTUs / FTUs upload remote monitoring data ("three-remote" data) to the power distribution automation master station (SCADA system) via another network. The two systems are physically or logically isolated. The video system serves only as a security tool and has no data interaction with the electrical quantity monitoring system. When the line current exceeds the limit, the SCADA master station provides an alarm function, requiring maintenance personnel to manually switch to the video system to locate the camera and observe the footage.

[0004] In the process of developing the relevant technology, the applicant discovered that the relevant technology has at least the following technical problems: The problem of data silos is prominent. Video and electrical quantity data cannot be automatically correlated due to differences in networks, protocols, and platforms, resulting in fragmented monitoring information. Furthermore, video monitoring systems in related technologies can only assist manual verification of events that have already occurred. They cannot proactively issue warnings in the early stages of equipment anomalies (such as when meter pointers approach the red line). Moreover, the diagnosis of power distribution networks relies on cloud processing, resulting in poor real-time performance and an inability to effectively support the status monitoring of power equipment. Summary of the Invention

[0005] In view of this, this application provides a power distribution terminal device, power distribution network inspection method, device and medium based on edge computing and multimodal perception. The main purpose is to solve the problems of fragmented monitoring information, poor real-time performance and inability to effectively support the status monitoring of power equipment.

[0006] According to a first aspect of this application, a power distribution terminal device based on edge computing and multimodal sensing is provided, the power distribution terminal device including a control module and an Internet of Things module: The control module periodically acquires telemetry data from the power distribution network and transmits the telemetry data to the IoT module via a preset private protocol. The IoT module receives and stores the telemetry data transmitted by the control component, and receives real-time images of the power distribution network collected and uploaded by an externally connected camera component. It analyzes and processes the real-time images, generates inspection results of the power distribution network in combination with the telemetry data, and transmits the inspection results to the information management area for processing.

[0007] According to a second aspect of this application, a method for inspecting a power distribution network is provided, the method being applied to power distribution terminal equipment, comprising: Regularly acquire and store telemetry data from the power distribution network; The system receives and uploads real-time images of the power distribution network from externally connected camera components. The real-time images are analyzed and processed, and the inspection results of the power distribution network are generated by combining the telemetry data. The inspection results are then transmitted to the information management area for further processing.

[0008] According to a third aspect of this application, an apparatus is provided, including a memory and a processor, the memory storing a computer program, the processor executing the computer program to implement the steps of the method described in any of the second aspects above.

[0009] According to a fourth aspect of this application, a medium is provided having a computer program stored thereon, which, when executed by a processor, implements the steps of the method described in any one of the second aspects above.

[0010] By utilizing the above technical solutions, this application provides a power distribution terminal device, power distribution network inspection method, equipment, and medium based on edge computing and multimodal perception. This application embeds judgment capabilities into the power distribution terminal device, processes images collected by camera components in real time at the device edge, and achieves rapid intelligent recognition. While avoiding fragmented monitoring information, it couples power business logic and computer vision algorithms at the software level, enabling proactive early warning through edge computing capabilities in the early stages of equipment anomalies, without relying on cloud-based diagnostic processing. It has good real-time performance and reliability, ensuring effective support for the status monitoring of power equipment.

[0011] The above description is only an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, the following are specific embodiments of this application. Attached Figure Description

[0012] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the scope of this application. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings: Figure 1 This illustration shows a schematic diagram of the architecture of a power distribution terminal device based on edge computing and multimodal sensing, provided in an embodiment of this application. Figure 2 This illustration shows a real-time image cropping method provided in an embodiment of this application. Figure 3 This illustration shows a digital cutting schematic diagram provided by an embodiment of this application; Figure 4 This illustration shows a digital image segmentation diagram provided in an embodiment of this application; Figure 5 This illustration shows a flowchart of a real-time image recognition process to obtain current atlas data, according to an embodiment of this application. Figure 6 A schematic diagram of a comparative diagnosis process provided in an embodiment of this application is shown; Figure 7 This illustration shows a schematic diagram of another power distribution terminal device based on edge computing and multimodal sensing provided in an embodiment of this application; Figure 8 This paper illustrates a schematic flowchart of a power distribution network inspection method provided in an embodiment of this application. Figure 9 A schematic diagram of the device structure of a computer device provided in an embodiment of this application is shown. Detailed Implementation

[0013] Exemplary embodiments of the present application will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present application are shown in the drawings, it should be understood that the present application may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this application will be thorough and complete, and will fully convey the scope of the present application to those skilled in the art.

[0014] This application provides a power distribution terminal device based on edge computing and multimodal sensing, such as... Figure 1 As shown, the power distribution terminal equipment includes a control module 101 and an IoT module 102. The control module 101 periodically acquires telemetry data from the power distribution network through its built-in telemetry function module. This telemetry data includes, but is not limited to, electrical parameters such as current, voltage, and power, as well as status variables such as switch states. In this embodiment, the control module 101 follows the functional specifications of existing power distribution terminals and uploads the acquired telemetry data to the IoT module via a dedicated power communication network (such as fiber optic Ethernet) or a wireless network using power protocols (such as IEC 60870-5-104). During the upload of telemetry data to the IoT module 102, the control module 101 and the IoT module 102 communicate using a preset proprietary protocol. This proprietary protocol is a communication protocol designed for this power distribution terminal device, aiming to optimize data transmission efficiency and enhance system security, ensuring both high efficiency and security of data transmission.

[0015] The IoT module 102 receives and stores telemetry data from the control module 101 for subsequent analysis. Simultaneously, the IoT module 102 receives real-time images of the power distribution network collected and uploaded by externally connected camera components via standard interfaces such as RS485. These camera components include, but are not limited to, visible light cameras and infrared thermal imaging cameras, used to capture images of equipment appearance and temperature data. Furthermore, the IoT module 102 incorporates an AI computing core equipped with infrared image recognition technology based on the KNN machine learning algorithm, enabling intelligent analysis and processing of the received real-time images. Specifically, the AI ​​computing core first preprocesses the images, including cropping, denoising, and binarization, then extracts image feature values ​​and compares them with preset training samples to achieve intelligent recognition of image content. Combining the telemetry data received from the control module 101, the IoT module 102 generates detailed inspection results of the power distribution network and uploads these results to the information management area for maintenance personnel to reference.

[0016] The IoT module 102 includes a feeder terminal unit agent module 1021, an analysis and application module 1022, a data bus module 1023, and a northbound application module 1024. The feeder terminal unit proxy module 1021 and the control module 101 establish a data link through a unidirectional isolation module. This unidirectional isolation module is a hardware or software-implemented isolation mechanism that ensures data can only flow in one direction, enhancing system security and preventing external interference from affecting the control module 101. Based on this data link, the feeder terminal unit proxy module 1021 receives telemetry data transmitted from the control module 101, performs protocol conversion on the telemetry data, transforming it into a format recognizable by the data bus module 1023, and then transmits the converted telemetry data to the data bus module 1023 for further processing.

[0017] The analysis and application module 1022 is the intelligent analysis core of the IoT module 102. It can receive real-time images of the power distribution network collected and uploaded by camera components (including visible light cameras and infrared thermal imaging cameras). The analysis and application module 1022 has a built-in image recognition engine based on the KNN algorithm. The KNN algorithm is an instance-based learning method that classifies or regresses by comparing the similarity between new instances and instances in the training set. This engine first preprocesses the images, including cropping, denoising, and binarization, to extract key features. Then, based on the KNN algorithm, it recognizes the real-time images to obtain the current map data, such as equipment temperature distribution and appearance anomalies. Furthermore, the analysis and application module 1022 compares the current map data with historical map data to identify subtle changes in equipment status, obtains the analysis results, and transmits the analysis results to the data bus module 1023.

[0018] The data bus module 1023 receives and stores telemetry data transmitted by the feeder terminal unit agent module 1021, and receives and stores the analysis results transmitted by the analysis and application module 1022. Through an efficient data storage and management mechanism, it ensures the integrity and accessibility of the data. After receiving both types of data, the data bus module 1023 combines the telemetry data and analysis results to generate inspection results. These results may include equipment electrical parameter information, image recognition results, and anomaly warning information, providing maintenance personnel with a comprehensive overview of the equipment status. The inspection results are then transmitted to the northbound application module 1024 for further processing.

[0019] The northbound application module 1024 is used to provide data or services to the upper-layer systems or applications of the IoT module. Specifically, the northbound application module 1024 receives inspection results and transmits them to the information management area for processing. In actual application, the information management area can be Zone IV, which undertakes data aggregation and upper-layer management functions, supports power equipment status monitoring and intelligent early warning, and reflects the collaborative operation of edge computing and cloud management.

[0020] Thus, through the aforementioned IoT module 102 architecture, real-time image processing and intelligent recognition are achieved at the edge of the power distribution terminal equipment, avoiding fragmented monitoring information and improving the real-time performance and reliability of data processing. Furthermore, by coupling power business logic with computer vision algorithms at the software level, proactive early warnings can be issued through edge computing capabilities at the initial stage of equipment anomalies, without relying on cloud-based diagnostic processing, thereby effectively supporting the status monitoring and operation and maintenance management of power equipment. For example, in a power distribution room inspection, the feeder terminal unit agent module successfully received telemetry data from the control module, including the current and voltage values ​​of a certain switchgear. Simultaneously, the analysis application module captured real-time images and infrared thermal images of the switchgear through the camera component. Using the built-in KNN algorithm, the module intelligently recognized the images and discovered an abnormal temperature rise in a certain part of the switchgear. Subsequently, the module compared the current image data with historical image data for diagnosis, confirming that the anomaly was a potential fault. The data bus module combined the telemetry data and analysis results to generate a detailed inspection report, which was then transmitted to the power distribution automation master station through the northbound application module. After receiving the warning, the maintenance personnel quickly located the faulty switchgear and took corresponding repair measures, effectively preventing equipment failure and ensuring the safe and stable operation of the power distribution network.

[0021] As described above regarding the feeder terminal unit proxy module 1021, the feeder terminal unit proxy module 1021 performs protocol conversion on telemetry data and restructures multi-source data according to power communication standards. The specific conversion process is as follows: First, basic parameter data (such as terminal ID, communication parameters, etc.) is extracted from the telemetry data and labeled using a common logic node LD (inst named "LD0") to achieve a unified description of device identity and communication attributes. Next, measured values ​​such as current and voltage are extracted from the telemetry data and encapsulated using a secondary IED telemetry logic node IMEAS to form measurement data blocks with time stamp and dimensional attributes. Control command data such as switch control and voltage regulation commands are encapsulated using a secondary IED control logic node ICTRL to ensure the complete timing and validity identification of control commands. Finally, status data such as device operating status and anomaly flags are extracted from the telemetry data and encapsulated using a secondary IED status logic node IBRST to form a status matrix that can be parsed by the upper-layer system, thus completing the protocol conversion of the telemetry data. For example, the raw data uploaded by a switchgear terminal includes the device number, A-phase current value, remote tripping command, and overcurrent alarm status. After conversion, LD0 labels the device identity information, IMEAS encapsulates the current measurement value (including time stamp and unit), ICTRL encapsulates the tripping command (including execution timing), and IBRST encapsulates the overcurrent status flag. This conversion mechanism achieves unified mapping of data semantics through standardized logical nodes, enabling seamless parsing of data from different manufacturers' devices in IoT modules. This improves data compatibility and processing efficiency at the edge, providing a structured data foundation for rapid diagnosis of subsequent application modules and effectively supporting edge early warning and status monitoring in the early stages of power equipment anomalies.

[0022] As described above regarding the analysis and application module 1022, the analysis and application module 1022 uses the KNN algorithm to identify real-time images and obtain the current map data. The specific identification process is as follows: In practice, the infrared thermal imager carried by the robot captures a combination of an infrared thermogram and a daylight image. Since only the infrared thermogram can display temperature, the images in the folder are categorized based on the pixel values ​​of the infrared thermogram and the daylight image during input. The infrared thermogram image with a pixel value of 640×512 is selected as the real-time image for recognition. A preset key region mapping technique is used to locate the position of the infrared thermogram within the mixed image. The infrared thermogram image captured by the robot is then cropped to ensure that only the effective temperature measurement area is retained. Specifically, the highest temperature display in the infrared thermogram image is always in the upper left corner of the image; therefore, only the upper left corner is retained. Subsequently, the cropped real-time image is binarized to obtain a binarized image. In each image, each pixel represents a different color, and in a computer, each color is represented by a different grayscale number; for example, the number 1 represents black, and the number 255 represents white. Therefore, after obtaining the cropped real-time image, it can be de-noised and binarized. Because the grayscale values ​​in the cropped image are quite extreme, only grayscale values ​​greater than or equal to 150 are assigned 1, and the rest are assigned 0. For example, as shown... Figure 2 As shown, the cropped real-time image is a grayscale image of the number "6". Converting it... Figure 2 The image shows a binary matrix consisting of 0s and 1s.

[0023] Given the characteristics of multi-digit displays in binarized images, considering that a multi-digit number can be composed of multiple unit digits (e.g., 11.6 can be composed of the digits 1 and 6, plus a decimal point), and that the smallest identifiable digit is a unit digit within 10, it is necessary to segment the data in the image and then identify each digit individually. The segmentation process is essentially the process of partitioning the binary matrix. Finding the transition points between 0 and 1 and between 1 and 0 allows us to determine the location of the segmentation point. This means identifying at least one transition point in the binarized image and segmenting the image according to at least one transition point to obtain multiple digit images. For example, see [link to example]. Figure 3 The number "11.6" shown is segmented by detecting grayscale transition points from 0 to 1 and from 1 to 0. "11.6" is first vertically segmented into three character regions, and then the decimal point is separated by horizontal segmentation to finally obtain a single digit image block, thus solving the recognition problem caused by the continuous display of multiple digits.

[0024] After image preprocessing, features of each digital image need to be extracted before subsequent processing can proceed. Specifically, the following processing needs to be performed on each digital image: Divide the digital image into multiple regions according to a preset segmentation standard. For example, see... Figure 4This method can segment a digital image into nine regions, A1 to A9. Then, the percentage of a specified value in each region is calculated, resulting in multiple percentage values. Specifically, the percentage of '1' in each region can be selected as the percentage value, which indicates the image feature. (See also...) Figure 4 Calculate the proportion of the area occupied by the number 1 in A1, A2, A3, A4, A5, A6, A7, A8, and A9 of the image. Then calculate the proportion A10 of all 1s in the entire image. This extracts the feature vector A = [A1, A2, A3, A4, A5, A6, A7, A8, A9, A10]. Taking the numbers 1, 1, and 6 as examples, the calculated proportions are shown in Table 1 below. Table 1

[0025] Table 1 shows the feature value distribution of different regions (A1 to A10) for the corresponding digits 1 and 6. As shown in Table 1, there is a significant difference in feature values ​​between digit 1 and digit 6 within the same region. For example, the feature value of digit 1 in region A2 is close to 0, while the feature value of digit 6 is as high as 0.866667, demonstrating the clear distinguishability of feature representation for different digits within the same image region. The feature values ​​of the same digit also show a regular variation across different regions. For instance, the feature value of digit 1 reaches around 0.8 in regions A5 and A6, while approaching 0 in regions A2 and A7, reflecting the difference in the contribution of different spatial locations of the digit image to feature extraction. Therefore, based on the recognition results in Table 1, using the digits 1, 1, and 6 as image feature values ​​is feasible. These feature values ​​are generated by statistically analyzing the distribution ratio of binarized pixels in each region, providing a key basis for the KNN algorithm to distinguish different digits. By calculating the Euclidean distance between the input image feature vector and the sample features, accurate digit classification is achieved. It should be noted that, in this embodiment of the application, 10 image feature values ​​are selected for an image. Therefore, the overall calculation can be performed in the form of a matrix to reduce the space complexity of the code.

[0026] In this way, through the above process, the image feature value corresponding to the digital image can be determined according to the multiple numerical ratios. After calculating the feature value of the image, it is necessary to compare and analyze the feature values of different values with the feature values in the training samples to complete digital classification. Specifically, it is necessary to first read the training results of the pre-trained samples. In the embodiment of the present application, 5 images intercepted from 0 to 9, ".", and the character "摄" can be placed in 12 folders named 0 to 11 respectively (where 0 to 9 and "." are used for digital recognition in the image, and the character "摄" is used as the termination feature of the entire recognition program), and then the algorithm is trained through a function to obtain the training results of the samples and save the training results of the samples in the train_set.pkl file. In this way, the training results of the samples can be read from this file.

[0027] Next, using the KNN algorithm, calculate the Euclidean distance between the image feature value of each digital image and each feature value in the sample training results to determine a single digit corresponding to each digital image, and obtain multiple single digits corresponding to multiple digital images. Among them, the KNN algorithm is the core. After effectively training the data model, when new data is input, the corresponding output quantity is mapped through the model. Under a fixed training set, the three important factors of the KNN algorithm are the selection of the k value, the distance measurement method, and the classification decision rule. Among them, the classification decision rule uses the majority voting method, and for the selection of the k value and the distance measurement method, the present application executes according to the following strategy: First, for the selection of the k value, generally, a smaller value needs to be randomly selected according to the distribution of the samples, or a more appropriate k value can be selected by the method of cross-validation. If a relatively small k value is selected, it means that training instances are used for prediction in a smaller domain, and the training error will be relatively small. Generally, the prediction result will only take effect when the input instance is similar or consistent with the training instance. However, at the same time, the generalization error will also increase, that is, the reduction of the k value will make the overall model more complex and prone to overfitting; if a larger k value is selected, it is equivalent to using training instances in a larger domain for prediction. Although the generalization error can be reduced, the training error will also increase at the same time. At this time, training instances that are not very similar to the input instance will also have a greater impact on the predicted value, resulting in a decrease in the prediction accuracy, but the increase of the k value will make the overall model simpler. Therefore, in the embodiment of the present application, the k value is finally selected as 20 through the method of multiple cross-validations.

[0028] Secondly, for the distance measurement method, the embodiment of the present application adopts the Euclidean distance. For two n-dimensional vectors x and y, the Euclidean distance between the two can be expressed by the following formula 1: Formula 1:

[0029] In Formula 1, and They are two n-dimensional vectors for which distance calculation is to be performed. , Each component of a vector corresponds to a coordinate value in n-dimensional space; Representing vectors and The Euclidean distance between them.

[0030] Thus, through the above process, Euclidean distance is selected as the feature similarity metric, and digital classification is achieved by calculating the Euclidean distance between the input image feature values ​​and the sample features.

[0031] Finally, the identified individual numbers are integrated as floating-point data, such as combining "1", "1", and "6" into 11.6, to obtain the current spectrum data and output it. The analysis and application module 1022 is also used to save the current spectrum data in dictionary form for subsequent diagnostic use. In this way, by combining local image features with the KNN algorithm, the problems of complex backgrounds and diverse temperature value display forms in infrared images of power equipment are effectively overcome. Millisecond-level real-time identification is achieved at the edge, ensuring rapid capture and accurate recording of abnormal temperatures of equipment, and significantly improving the intelligent monitoring level of power distribution terminals.

[0032] In summary, the process of identifying real-time images to obtain the current atlas data is summarized as follows: Figure 5 As shown, firstly, a sample training set is input, image features are extracted from the sample training set, and the algorithm is trained using the sample training set. The training results are then saved to the `train_set.pkl` file. When new data (real-time acquired infrared images of the device) is input, the images are preprocessed and digitally segmented. Image feature values ​​of each digit image are extracted, and the Euclidean distance between the digit and the feature vectors of each category in the training results saved in `train_set.pkl` is calculated using the KNN algorithm. The nearest neighbor category is determined based on the majority voting method, thereby identifying individual digits. Finally, multiple identified digits are integrated into floating-point map data according to the image display order (e.g., the segmented and identified "1", "1", ".", and "6" are integrated into 11.6), forming a quantitative output result that reflects the temperature status of the device, realizing rapid digit recognition and map data generation at the edge.

[0033] As described above regarding the analysis and application module 1022, the analysis and application module 1022 will compare and diagnose the current map data with historical map data to obtain the analysis results. The specific comparison and diagnosis process is as follows: First, historical temperature data is acquired, including historical temperature values ​​and timestamps. By calculating the absolute difference between the current and historical data, the magnitude of temperature change and key point differences are determined. For example, if the current temperature measurement at a switchgear A-phase connection point is 78.5℃, while the historical data from three days ago is 72.3℃, a difference of 6.2℃ is automatically marked. Then, the current and historical data are integrated into a time-series table containing fields such as "collection time," "temperature value," and "rate of change." Areas exceeding the difference limit are highlighted with color (e.g., a temperature difference exceeding 5℃ is marked in red), generating a judgment result with difference annotations. This process achieves visualized tracking of equipment status changes through quantitative comparison, enabling timely detection of abnormal temperature rise trends. For example, if a device shows a temperature increase after three consecutive inspections with a single temperature difference exceeding 3℃, an alert can be automatically triggered and pushed to the maintenance terminal. This diagnostic mechanism based on historical data comparison effectively solves the problem of random errors in single temperature measurements, improving the accuracy of equipment status assessment and the timeliness of fault warnings.

[0034] In summary, the comparative diagnostic process can be summarized as follows: Figure 6 As shown, the historical temperature measurement data is parsed during the process, and the parsed historical data is stored in the `last_data_set.pkl` file as a dictionary. When new current temperature measurement data is detected, the existing historical data in this file is read, and the current temperature measurement data is added to the dictionary value set. The integrated data is then output as an Excel spreadsheet. During this process, the data from the two most recent collections are compared and analyzed. Finally, the diagnostic results are filled into the last column of the spreadsheet, forming a complete report that includes historical data comparisons and diagnostic conclusions.

[0035] As described above regarding the analysis application module 1022, it transmits the analysis results to the data bus module 1023. Specifically, when transmitting the analysis results to the data bus module 1023, the analysis application module 1022 follows a structured message construction process to ensure the standardization and reliability of data transmission: First, it determines the preset message format, which can be shown in Table 2 below. Table 2

[0036] As shown in Table 2, the preset message format employs a structured data transmission framework. Using "68H 68H 68H" as a fixed starting identifier ensures the receiver can quickly identify message boundaries. Subsequently, a dynamic numbering mechanism is established using data sequence numbers, employing a decimal cyclic counting method (00H to 09H cyclically increasing) to achieve sequential management of data packets, effectively avoiding number overflow issues in long-sequence transmissions. The core data field contains zero-sequence voltage waveform data and zero-sequence current waveform data, each 6000 bytes long, providing ample storage space for power system waveform sampling and meeting the requirements of high-precision time-series data transmission. The check character is formed by summing the original data bytes of zero-sequence voltage and current and truncating the lower 8 bits. This algorithm effectively detects data distortion during transmission while ensuring computational efficiency. Finally, the message is encapsulated with "16H 16H 16H" as the end identifier, forming a complete data encapsulation structure that includes a start identifier, ordered data packets, integrity check, and end identifier. This design improves parsing efficiency through fixed identifiers, optimizes data packet management through a cyclic sequence number mechanism, and enhances transmission reliability through check and verification. The overall solution takes into account the real-time, accuracy, and anti-interference requirements of power system data transmission.

[0037] Based on the aforementioned preset message format, a blank message starting with the message start identifier "68H 68H68H" is created. Subsequently, the analysis results are split into data, and each split data item is assigned an incrementing sequence number (the first item is numbered 00H, the tenth is 09H, and after ten items, the sequence number is reset to 00H). This sequence is then appended to the blank message after the message start identifier, ensuring the sequential order of data packets and avoiding number overflow issues caused by long sequences. For example, when transmitting 12 data items, the first 10 items are numbered sequentially from 00H to 09H, and the eleventh item is numbered starting again from 00H. Simultaneously, zero-sequence voltage waveform data and zero-sequence current waveform data are extracted from the analysis results and sequentially filled into the core data field of the blank message. Both are preset to have a capacity of 6000 bytes to accommodate typical waveform sampling scales; for example, voltage and current timing data from a fault recording can be completely stored here. After data filling, the original byte values ​​of zero-sequence voltage and current are added point by point, and the lower 8 bits are taken to calculate the check character (e.g., if the data sum is 0x1A3F, then the check character is 0x3F). This can effectively detect single-bit errors or multi-bit overflow errors during transmission and fill them into the blank message. A preset end marker "16H 16H 16H" is added to the end of the blank message to obtain the target message. The target message is then transmitted to the data bus module. In this way, parsing efficiency is improved by fixing the message header and footer identifiers, data packet management is optimized by the sequence number cycle mechanism, and data integrity is enhanced by checksum verification. The overall process takes into account the real-time and accuracy requirements of high-frequency sampling data transmission in the power system, and significantly reduces the risk of data misordering, loss, or tampering.

[0038] In another alternative implementation, the IoT module 102 further includes a data center module 1025: When the generated assessment result indicates a fault, the assessment application module 1022 utilizes real-time images (recording the device's appearance at the time of the fault), current map data (marking abnormal temperature values ​​and collection timestamps), and inspection results (including historical device operating parameters and current alarm types) to generate a fault time file, and transmits the fault time file to the data bus module 1023; the data bus module 1023 receives the fault time file and transmits it to the data center module; the data center module receives the fault... The time-lapse files can be stored according to a composite index rule of "device ID-fault type-occurrence time". At the same time, a correlation relationship with historical inspection data can be established. For example, the overheating fault file of a phase A connector of a switch cabinet can be automatically associated with the temperature curve data of the equipment over the past three months, realizing the full-chain retention of fault evidence. When maintenance personnel conduct subsequent analysis, they can simultaneously retrieve real-time images (showing burn marks on the connector), spectral data (78.5℃ over-temperature value), and historical temperature trends (temperature rise curves for three consecutive weeks) at the time of the fault. This effectively solves the limitation of traditional methods that can only provide a single data source, and significantly improves the accuracy and processing efficiency of fault cause analysis.

[0039] In another alternative implementation, the internal communication of the IoT module 102 actually adopts the DDS communication protocol, which can be implemented in code. The code logic is as follows: Based on the cffi toolkit, the prototype of the C interface function is determined and the callback type of the C interface function prototype is determined. According to the callback type, the callback processing function used to convert C language format data into Python string is determined. When the IoT model is running, the language communication plugin matching the current system type is started, and the communication bus is started. The communication bus is used to publish JSON format data of the specified topic to the outside world. At the same time, the target topic is subscribed to and the callback processing function is registered to receive the subscription information fed back to the subscribed topic and use the callback processing function to convert the format of the subscription information.

[0040] As described above, the IoT module provides a message subscription function. The specific details of the topic configuration and data item definitions used for uploading data information during failures can be found in Table 3 below: Table 3

[0041] As shown in Table 1, the topics are actively pushed by the analysis and application module, and the receiver adopts an on-demand subscription model to receive the data. Its main purpose is to achieve real-time uploading and sharing of information related to equipment failures. The data item definition includes three core attributes: Fault Signal (FltStau) uses an integer data type, identifying changes in the equipment failure state through two status values: 0 (reset) and 1 (action); Fault Time (FltTm) is recorded in string format with nanosecond-level time precision to ensure the accuracy of the failure occurrence time; Fault Type (FltTyp) uses an integer numerical encoding, where 0 indicates that the cause of the failure cannot be determined, and 1 indicates that a personnel intrusion event has been detected. This data structure, through the combination of status identifiers, precise timestamps, and type encoding, provides standardized data support for rapid fault event location, cause tracing, and emergency response, and is particularly suitable for time- and status-sensitive scenarios such as power equipment anomaly monitoring. Therefore, when the analysis application module 1022 in the IoT module 102 receives the subscription information, it determines the recipient of the subscription information, generates topic information including fault information, fault time and fault type, and pushes the topic information to the recipient. The fault information is based on 0 or 1 to indicate whether the fault has occurred, the fault time is used to indicate the time when the fault occurred, and the fault type is based on 0 or 1 to indicate human intrusion or other reasons when the fault occurred.

[0042] In another alternative implementation, when the control module 101 transmits telemetry data to the IoT module 102, it first determines a preset message format, which may be as shown in Table 4 below: Table 4

[0043] Table 4 defines a hexadecimal data packet structure, whose overall framework consists of six sequentially arranged fields: The data packet uses a double start character design to enhance synchronization reliability; both the header and the third field are 1-byte start characters (EB), forming a characteristic identifier; the following 2-byte length field indicates the data field capacity, providing a size basis for variable-length data transmission; the middle part is the data field, represented by a grid pattern, whose content can carry different information according to actual needs; subsequently, a 1-byte checksum field verifies data integrity through an accumulation algorithm (ignoring overflow bits); finally, a 1-byte stop character (D7) serves as the packet termination identifier. This structural design, with its double start characters strengthening frame header identification, dynamic length field adaptation to data capacity, and check mechanism ensuring transmission reliability, is suitable for communication scenarios requiring strict data encapsulation and error detection. Therefore, the control module 101 sets a blank message according to the preset message format, fills the data field of the blank message with telemetry data, calculates the length of the data field and fills the calculated result into the length field of the blank message, calculates the checksum for the telemetry data and appends the checksum to the end of the telemetry data in the blank message according to the preset message format, adds a start character, a repeat start character, and a stop character to the blank message according to the preset message format to obtain the target message, and transmits the target message to the IoT module 102. The IoT module 102 receives the target message transmitted by the control module 101, scans the target message, determines the repeat start character and stop character, truncates the target message according to the repeat start character and stop character, calculates the checksum of the truncated target message, and if the calculated result is the same as the checksum carried in the target message, the truncated target message is used as the telemetry data.

[0044] In summary, the architecture of the power distribution terminal equipment proposed in this application is summarized as follows: See Figure 7The system includes a control module 101 and an IoT module 102. The control module 101 serves as the external control source, and secure data interaction is achieved between the control module 101 and the feeder terminal unit proxy module 1021 in the IoT module 102 via a unidirectional isolation module, ensuring the secure transmission of control commands and telemetry data. The feeder terminal unit proxy module 1021 transmits the collected telemetry data via the data bus module 1023, supporting equipment status monitoring and information forwarding. Furthermore, the IoT module 102 integrates a northbound application module 1024 and a judgment application module 1022. The former handles routine business logic, while the latter uses real-time images collected by the camera components in the information management area, combined with the KNN algorithm, to perform infrared spectrum recognition and fault judgment. The generated judgment results are fed back to the northbound application module 1024 via the data bus module 1023 and uploaded to the information management area. Simultaneously, the image file at the time of the fault is synchronously transmitted to the data center module 1025 for storage. Through the collaboration of these modules, the secure issuance of control commands and real-time monitoring of equipment status are ensured, while also achieving the organic integration of intelligent image diagnosis and historical data tracing.

[0045] The power distribution terminal equipment provided in this application embeds analytical capabilities into the power distribution terminal equipment, processes images collected by the camera components in real time at the edge of the equipment, and achieves rapid intelligent identification. While avoiding the fragmentation of monitoring information, it couples power business logic with computer vision algorithms at the software level, enabling proactive early warning through edge computing capabilities in the early stages of equipment anomalies, without relying on cloud-based diagnostic processing. It has good real-time performance and reliability, ensuring effective support for the status monitoring of power equipment.

[0046] Furthermore, embodiments of this application provide a distribution network inspection method, such as... Figure 8 As shown, this method is applied to power distribution terminal equipment, including: S10: Periodically acquire and store telemetry data from the distribution network.

[0047] The specific process of periodically acquiring and storing telemetry data from the distribution network can be found in the description of this part in the previous embodiment, and will not be repeated here.

[0048] S20: Receives and uploads real-time images of the power distribution network from externally connected camera components.

[0049] The specific process of receiving and uploading real-time images of the power distribution network from externally accessed camera components can be found in the description of this part in the previous embodiment, and will not be repeated here.

[0050] S30: Analyze and process real-time images, combine them with telemetry data to generate inspection results for the power distribution network, and transmit the inspection results to the information management area for processing.

[0051] The specific process of analyzing and processing real-time images, generating inspection results of the power distribution network by combining telemetry data, and transmitting the inspection results to the information management area for processing can be found in the description of this part in the previous embodiment, and will not be repeated here.

[0052] Optionally, step S30, namely, the analysis and processing of the real-time image and the generation of the inspection result of the power distribution network in combination with the telemetry data, includes: Based on the KNN algorithm, the real-time image is identified to obtain the current map data; the current map data is compared and diagnosed with historical map data to obtain the judgment result; the inspection result is generated by combining the telemetry data and the judgment result. A detailed description of this process can be found in the description of this part in the previous embodiment, and will not be repeated here.

[0053] Optionally, in step S30, that is, based on the KNN algorithm, the real-time image is identified to obtain the current map data, including: The method involves determining a preset key region, mapping this region onto a real-time image, cropping the real-time image, and binarizing the cropped image to obtain a binarized image. At least one transition point is identified in the binarized image, and the image is segmented according to this transition point to obtain multiple digital images. For each digital image, the following processing is performed: the image is divided into multiple regions according to a preset segmentation standard, and the percentage of a specified value in each region is calculated to obtain multiple percentage values. Based on these percentage values, the corresponding image feature value is determined. The method also involves reading the pre-trained sample training results, using the KNN algorithm to calculate the Euclidean distance between the image feature value of each digital image and each feature value in the sample training results to determine the individual digit corresponding to each digital image, resulting in multiple individual digits corresponding to multiple digital images. These individual digits are then integrated as floating-point data to obtain the current atlas data and output. The method further includes saving the current atlas data in dictionary form. For a detailed description of this process, please refer to the description of this part in the previous embodiment; it will not be repeated here.

[0054] Optionally, in step S30, the current map data is compared and diagnosed with historical map data to obtain the judgment results, including: Historical map data is acquired, and the data differences between the current map data and the historical map data are identified. The current map data and the historical map data are integrated to generate a map data table. The map data table is then labeled using the data differences to obtain the analysis results. A detailed description of this process can be found in the description of this part in the previous embodiment, and will not be repeated here.

[0055] Optionally, the method further includes: If the assessment results indicate a fault, a fault time file is generated using real-time images, current map data, and inspection results; this fault time file is then stored. A detailed description of this process can be found in the description of this part in the previous embodiment, and will not be repeated here.

[0056] The method provided in this application embodiment periodically acquires and stores telemetry data of the power distribution network, receives real-time images of the power distribution network collected and uploaded by externally accessed camera components, analyzes and processes the real-time images, generates inspection results of the power distribution network in combination with telemetry data, and transmits the inspection results to the information management area for processing. By embedding the analysis capability into the power distribution terminal equipment, the images collected by the camera components are processed in real time at the edge of the equipment, realizing rapid intelligent identification. While avoiding the fragmentation of monitoring information, the method couples the power business logic and computer vision algorithms at the software level, enabling proactive early warning through edge computing capabilities in the early stages of equipment anomalies, without relying on cloud-based diagnostic processing. It has good real-time performance and reliability, ensuring effective support for the status monitoring of power equipment.

[0057] It should be noted that other corresponding descriptions of the steps involved in the power distribution network inspection method provided in this application embodiment can be found in the following references. Figures 1 to 7 The corresponding descriptions in [the document] will not be repeated here.

[0058] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties.

[0059] The above embodiments and the technical features in the embodiments can be combined arbitrarily. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0060] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

[0061] In an exemplary embodiment, see Figure 9The invention also provides a computer device including a bus, a processor, a memory, and a communication interface. It may also include an input / output interface and a display device, wherein the various functional units can communicate with each other via the bus. The memory stores a computer program, and the processor executes the program stored in the memory to perform the power distribution network inspection method described in the above embodiments.

[0062] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the power distribution network inspection method.

[0063] Through the above description of the embodiments, those skilled in the art can clearly understand that this application can be implemented in hardware or by using software plus necessary general-purpose hardware platforms. Based on this understanding, the technical solution of this application can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (such as a CD-ROM, USB flash drive, external hard drive, etc.) and includes several instructions to cause a computer device (such as a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments of this application.

[0064] Those skilled in the art will understand that the accompanying drawings are merely schematic diagrams of a preferred embodiment, and the modules or processes shown in the drawings are not necessarily essential for implementing this application.

[0065] Those skilled in the art will understand that the modules in the apparatus of the implementation scenario can be distributed within the apparatus of the implementation scenario as described, or they can be located in one or more apparatuses different from this implementation scenario, with corresponding changes. The modules of the above-described implementation scenario can be combined into one module, or they can be further divided into multiple sub-modules.

[0066] The serial numbers in this application are for descriptive purposes only and do not represent the superiority or inferiority of the implementation scenario.

[0067] The above disclosures are only a few specific implementation scenarios of this application. However, this application is not limited to these. Any variations that can be conceived by those skilled in the art should fall within the protection scope of this application.

Claims

1. A power distribution terminal device based on edge computing and multimodal sensing, characterized in that, The power distribution terminal equipment includes a control module and an IoT module. The control module periodically acquires telemetry data from the power distribution network and transmits the telemetry data to the IoT module via a preset private protocol. The IoT module receives and stores the telemetry data transmitted by the control component, and receives real-time images of the power distribution network collected and uploaded by an externally connected camera component. It analyzes and processes the real-time images, generates inspection results of the power distribution network in combination with the telemetry data, and transmits the inspection results to the information management area for processing.

2. The power distribution terminal equipment according to claim 1, characterized in that, The IoT module includes a feeder terminal unit agent module, an analysis and application module, a data bus module, and a northbound application module. The feeder terminal unit agent module establishes a data link with the control module through a one-way isolation module. Based on the data link, it receives the telemetry data transmitted by the control module, performs protocol conversion on the telemetry data, and transmits the converted telemetry data to the data bus module. The analysis and application module receives real-time images of the power distribution network collected and uploaded by the camera component, identifies the real-time images based on the KNN algorithm to obtain the current map data, compares the current map data with historical map data to obtain the analysis result, and transmits the analysis result to the data bus module. The data bus module receives and stores the telemetry data transmitted by the feeder terminal unit agent module, receives and stores the judgment result transmitted by the judgment application module, and generates the inspection result by combining the telemetry data and the judgment result and transmits it to the northbound application module. The northbound application module receives the inspection results and transmits them to the information management area for processing. The northbound application module is used to provide data or services to the upper-layer system or application of the IoT module.

3. The power distribution terminal equipment according to claim 2, characterized in that, The feeder terminal unit proxy module performs protocol conversion on the telemetry data, including: Basic parameter data is extracted from the telemetry data and labeled using common logic nodes. Measurement value data is extracted from the telemetry data and encapsulated using telemetry logic nodes. Control command data is extracted from the telemetry data and encapsulated using control logic nodes. Status data is extracted from the telemetry data and encapsulated using status logic nodes to complete the protocol conversion of the telemetry data.

4. The power distribution terminal equipment according to claim 2, characterized in that, The analysis and application module, based on the KNN algorithm, identifies the real-time image to obtain the current image data, including: A preset key region is determined, the preset key region is mapped onto the real-time image and the real-time image is cropped, and the cropped real-time image is binarized to obtain a binarized image. Identify at least one transition point in the binarized image, and cut the binarized image according to the at least one transition point to obtain multiple digital images; The following processing is performed on each of the digital images: the digital image is divided into multiple regions according to a preset segmentation standard, and the proportion of a specified value in each region is calculated to obtain multiple proportions. Based on the multiple proportions, the image feature value corresponding to the digital image is determined. Read the pre-trained sample training results, use the KNN algorithm to calculate the Euclidean distance between the image feature value of each digital image and each feature value in the sample training results, so as to determine the single number corresponding to each digital image, and obtain the multiple single numbers corresponding to the multiple digital images; The multiple individual numbers are integrated as floating-point data to obtain the current graph data and output it. The analysis and application module is also used to save the current graph data in dictionary form.

5. The power distribution terminal equipment according to claim 2, characterized in that, The analysis and application module compares and diagnoses the current atlas data with historical atlas data to obtain the analysis results, including: Obtain the historical map data and determine the data differences between the current map data and the historical map data; The current map data and historical map data are integrated to generate a map data table. The map data table is then labeled using the data differences to obtain the judgment result.

6. The power distribution terminal equipment according to claim 2, characterized in that, When the analysis application module transmits the analysis result to the data bus module, it creates a blank message starting with a message start identifier according to a preset message format, splits the analysis result into data, assigns an incrementing sequence number to each piece of data obtained from the splitting, arranges them, and appends them to the blank message after the message start identifier. It extracts zero-sequence voltage waveform data and zero-sequence current waveform data from the analysis result and fills them into the core data field of the blank message in sequence. It calculates check characters for the zero-sequence voltage waveform data and the zero-sequence current waveform data and fills them into the blank message. Finally, it adds a preset end marker to the end of the blank message to obtain the target message, and transmits the target message to the data bus module.

7. The power distribution terminal equipment according to claim 2, characterized in that, The IoT module also includes a data center module: If the generated assessment results indicate a fault, the assessment application module uses the real-time image, the current map data, and the inspection results to generate a fault time file and transmits the fault time file to the data bus module. The data bus module receives the fault time file and transmits the fault time file to the data center module; The data center module receives and stores the fault time file.

8. The power distribution terminal equipment according to claim 2, characterized in that, The internal communication of the IoT module adopts the DDS communication protocol. Based on the cffi toolkit, the prototype of the C interface function is determined and the callback type of the C interface function prototype is determined. According to the callback type, a callback processing function is determined to convert C language format data into Python string. When the IoT model is running, the language communication plugin matching the current system type is started, and the communication bus is started. The communication bus is used to publish JSON format data of the specified topic to the outside world. At the same time, the target topic is subscribed to and the callback processing function is registered to receive the subscription information fed back to the subscribed topic and use the callback processing function to convert the format of the subscription information.

9. The power distribution terminal equipment according to claim 8, characterized in that, When the analysis application module in the IoT module receives the subscription information, it determines the recipient that initiated the subscription information, generates topic information including fault information, fault time and fault type, and pushes the topic information to the recipient. The fault information is based on 0 or 1 to indicate whether a fault has occurred, the fault time is used to indicate the time when the fault occurred, and the fault type is based on 0 or 1 to indicate that the fault was caused by human intrusion or other reasons.

10. The power distribution terminal equipment according to claim 1, characterized in that, When the control module transmits the telemetry data to the IoT module, it sets a blank message according to a preset message format, fills the data field of the blank message with the telemetry data, calculates the length of the data field and fills the calculated result into the length field of the blank message, calculates a checksum for the telemetry data and appends the checksum to the blank message according to the preset message format, adds a start character, a repeat start character, and a stop character to the blank message according to the preset message format to obtain the target message, and transmits the target message to the IoT module. The IoT module receives the target message transmitted by the control module, scans the target message, determines the start repeat character and the stop repeat character, truncates the target message according to the start repeat character and the stop repeat character, calculates the checksum of the truncated target message, and if the calculated result is the same as the checksum carried in the target message, the truncated target message is used as the telemetry data.

11. A method for inspecting a power distribution network, characterized in that, The method is applied to power distribution terminal equipment, including: Regularly acquire and store telemetry data from the power distribution network; The system receives and uploads real-time images of the power distribution network from externally connected camera components. The real-time images are analyzed and processed, and the inspection results of the power distribution network are generated by combining the telemetry data. The inspection results are then transmitted to the information management area for further processing.

12. The method according to claim 11, characterized in that, The step of analyzing and processing the real-time image and generating the inspection result of the power distribution network in conjunction with the telemetry data includes: Based on the KNN algorithm, the real-time image is identified to obtain the current map data; The current map data is compared with historical map data to obtain the judgment result; The inspection results are generated by combining the telemetry data and the analysis results.

13. The method according to claim 12, characterized in that, The KNN algorithm is used to identify the real-time image to obtain the current image data, including: A preset key region is determined, the preset key region is mapped onto the real-time image and the real-time image is cropped, and the cropped real-time image is binarized to obtain a binarized image. Identify at least one transition point in the binarized image, and cut the binarized image according to the at least one transition point to obtain multiple digital images; The following processing is performed on each of the digital images: the digital image is divided into multiple regions according to a preset segmentation standard, and the proportion of a specified value in each region is calculated to obtain multiple proportions. Based on the multiple proportions, the image feature value corresponding to the digital image is determined. Read the pre-trained sample training results, use the KNN algorithm to calculate the Euclidean distance between the image feature value of each digital image and each feature value in the sample training results, so as to determine the single number corresponding to each digital image, and obtain the multiple single numbers corresponding to the multiple digital images; The multiple individual numbers are integrated as floating-point data to obtain the current graph data and then output. The method further includes saving the current atlas data in dictionary form.

14. The method according to claim 12, characterized in that, The step of comparing and diagnosing the current map data with historical map data to obtain the judgment result includes: Obtain the historical map data and determine the data differences between the current map data and the historical map data; The current map data and historical map data are integrated to generate a map data table. The map data table is then labeled using the data differences to obtain the judgment result.

15. The method according to claim 12, characterized in that, The method further includes: If the assessment results indicate a fault, a fault time file is generated using the real-time image, the current map data, and the inspection results. The fault time file is stored.

16. An apparatus comprising a memory and a processor, the memory storing a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 11 to 15.

17. A medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 11 to 15.

Images